Real‐Time Monitoring of Bubbles and Crashes

Given the financial and economic damage that can be caused by collapse of an asset price bubble, it is critical importance to rapidly detect onset a crash once bubble has been identified. We develop real-time monitoring procedure for detecting episode in time series. adopt autoregressive framework, with regimes modelled explosive stationary dynamics, respectively. The first stage our approach monitor bubble; conditional on which, we real as new data emerges. Our detection employs statistic based different signs means differences associated regimes, values are obtained using training period data. show desirable asymptotic properties terms its ability while never indicating earlier than one occurs. Monte Carlo simulations further demonstrate offer well-controlled false positive rate during regime. Application US housing market demonstrates efficacy house 2006.